knitr::opts_chunk$set(warning = FALSE)
source(here::here('code', 'helpers.R'))
library(tidyverse)
library(forcats)
library(cowplot)
##
## Attaching package: 'cowplot'
## The following object is masked from 'package:lubridate':
##
## stamp
library(agricolae)
library(ggupset)
library(RColorBrewer)
library(patchwork)
##
## Attaching package: 'patchwork'
## The following object is masked from 'package:cowplot':
##
## align_plots
library(vegan)
## Loading required package: permute
## Loading required package: lattice
## Registered S3 methods overwritten by 'vegan':
## method from
## plot.rda klaR
## predict.rda klaR
## print.rda klaR
## This is vegan 2.6-4
library(pheatmap)
library(ggradar)
library(broom)
data <- targets::tar_read(merged_all_results)
truth <- targets::tar_read(truth_set_data)
table(data$Type)
##
## BLAST100 BLAST97 CustomNBC Kraken_0.0 Kraken_0.05 Kraken_0.1
## 13668 20620 156465 105888 105888 105888
## Metabuli MMSeqs2_100 MMSeqs2_97 Mothur Qiime2 VSEARCH
## 75252 113184 113184 153171 22835 158112
Let’s remove the >0.2 Kraken runs, those are too strict
data <- data |> filter(!Type %in% c('Kraken_0.2', 'Kraken_0.3', 'Kraken_0.4', 'Kraken_0.5', 'Kraken_0.6', 'Kraken_0.7', 'Kraken_0.8', 'Kraken_0.9'))
#Made a mistake- Metabuli's and TNT's databases is misspelled
#data <- data |> mutate(Subject = str_replace_all(Subject, pattern = '_ref.fasta', replacement=''))
#data <- data |> mutate(Subject = str_replace_all(Subject, pattern = 'final.csv', replacement = 'final.fasta'))
Check whether any of the data is missing
type_list <- data |># filter(!str_detect(Subject, 'c01')) |>
group_by(Type) |>
summarize('Subject' = list(unique(Subject)))
all_subjects <- unique(data |># filter(!str_detect(Subject, 'c01')) |>
pull(Subject))
for(i in type_list$Type){
this_subjects <- type_list |> filter(Type == i) |> pull(Subject)
missing <- setdiff(all_subjects, this_subjects[[1]])
if (length(missing) > 0){
print(i)
print(missing)
}
}
## [1] "BLAST100"
## [1] "12S_v10_HmmCut.fasta" "16S_v04_HmmCut.fasta" "c01_v03_HmmCut.fasta"
## [1] "BLAST97"
## [1] "12S_v10_HmmCut.fasta" "16S_v04_HmmCut.fasta" "c01_v03_HmmCut.fasta"
## [1] "CustomNBC"
## [1] "c01_v03_HmmCut.fasta"
## [1] "Mothur"
## [1] "12S_v10_HmmCut.fasta" "16S_v04_HmmCut.fasta" "c01_v03_HmmCut.fasta"
OK, we can ignore the HmmCut ones.
data |> write_tsv('./results/cleaned_and_filtered_data.tsv.gz')
names(table(data$Query))
## [1] "KWest_16S_PooledTrue.fa"
## [2] "make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa"
## [3] "make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_noLulu_RESULTS_dada2_asv.fa"
## [4] "make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa"
## [5] "make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_noLulu_RESULTS_dada2_asv.fa"
## [6] "make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa"
## [7] "make_12s_16s_simulated_reads_6-fakeGenes_GreenGenes_RESULTS_dada2_asv.fa"
## [8] "make_12s_16s_simulated_reads_6-fakeGenes_Random_RESULTS_dada2_asv.fa"
## [9] "make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_12S_RESULTS_dada2_asv.fa"
## [10] "make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_16S_RESULTS_dada2_asv.fa"
## [11] "make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa"
## [12] "make_12s_16s_simulated_reads_8-Rottnest_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa"
## [13] "make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_12S_RESULTS_dada2_asv.fa"
## [14] "make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_16S_RESULTS_dada2_asv.fa"
table(data$Subject)
##
## 12s_v010_final.fasta
## 13640
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 13220
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 12886
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 12896
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 12979
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 13150
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 13251
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 12824
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 13015
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 13308
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 13010
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 13415
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 13322
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 13216
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 13386
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 13051
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 13344
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 12961
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 13118
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 13076
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 13330
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 12771
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 12440
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 12749
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 12845
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 12551
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 12682
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 12523
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 12566
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 12560
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 12603
## 12S_v10_HmmCut.fasta
## 10415
## 16S_v04_final.fasta
## 14732
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 13341
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 13417
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 13794
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 13546
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 13839
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 13666
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 13307
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 13899
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 13146
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 13261
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 14108
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 13846
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 14218
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 13889
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 14103
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 14427
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 14013
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 14083
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 14320
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 14126
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 13043
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 13388
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 13201
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 13250
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 13413
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 13230
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 13045
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 13120
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 12986
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 13093
## 16S_v04_HmmCut.fasta
## 11291
## c01_v03_final.fasta
## 9541
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 9209
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 9861
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 9256
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 9792
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 9085
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 9196
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 9180
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 9523
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 9120
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 9827
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 9361
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 9240
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 9277
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 9283
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 9326
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 9376
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 9216
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 9343
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 9424
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 9487
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 9899
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 9815
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 9708
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 9056
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 8997
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 9449
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 9719
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 9575
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 8975
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 8981
## c01_v03_HmmCut.fasta
## 5814
table(data$Subject)
##
## 12s_v010_final.fasta
## 13640
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 13220
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 12886
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 12896
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 12979
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 13150
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 13251
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 12824
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 13015
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 13308
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 13010
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 13415
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 13322
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 13216
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 13386
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 13051
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 13344
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 12961
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 13118
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 13076
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 13330
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 12771
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 12440
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 12749
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 12845
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 12551
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 12682
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 12523
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 12566
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 12560
## 12s_v010_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 12603
## 12S_v10_HmmCut.fasta
## 10415
## 16S_v04_final.fasta
## 14732
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 13341
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 13417
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 13794
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 13546
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 13839
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 13666
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 13307
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 13899
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 13146
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 13261
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 14108
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 13846
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 14218
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 13889
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 14103
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 14427
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 14013
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 14083
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 14320
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 14126
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 13043
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 13388
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 13201
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 13250
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 13413
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 13230
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 13045
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 13120
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 12986
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 13093
## 16S_v04_HmmCut.fasta
## 11291
## c01_v03_final.fasta
## 9541
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 9209
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 9861
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 9256
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 9792
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 9085
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 9196
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 9180
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 9523
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 9120
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 9827
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 9361
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 9240
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 9277
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 9283
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 9326
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 9376
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 9216
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 9343
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 9424
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 9487
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 9899
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 9815
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 9708
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 9056
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 8997
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 9449
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 9719
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 9575
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 8975
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 8981
## c01_v03_HmmCut.fasta
## 5814
twelveS_data <- data |> filter(Subject == '12s_v010_final.fasta')
sixteenS_data <- data |> filter(Subject == '16S_v04_final.fasta')
co1_data <- data |> filter(Subject == 'c01_v03_final.fasta')
table(twelveS_data$Query)
##
## KWest_16S_PooledTrue.fa
## 3174
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa
## 1123
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_noLulu_RESULTS_dada2_asv.fa
## 1123
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa
## 894
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_noLulu_RESULTS_dada2_asv.fa
## 894
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa
## 891
## make_12s_16s_simulated_reads_6-fakeGenes_GreenGenes_RESULTS_dada2_asv.fa
## 792
## make_12s_16s_simulated_reads_6-fakeGenes_Random_RESULTS_dada2_asv.fa
## 800
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_12S_RESULTS_dada2_asv.fa
## 268
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_16S_RESULTS_dada2_asv.fa
## 245
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa
## 243
## make_12s_16s_simulated_reads_8-Rottnest_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa
## 1053
## make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_12S_RESULTS_dada2_asv.fa
## 1139
## make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_16S_RESULTS_dada2_asv.fa
## 1001
table(sixteenS_data$Query)
##
## KWest_16S_PooledTrue.fa
## 3840
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa
## 1032
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_noLulu_RESULTS_dada2_asv.fa
## 996
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa
## 1108
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_noLulu_RESULTS_dada2_asv.fa
## 1108
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa
## 891
## make_12s_16s_simulated_reads_6-fakeGenes_GreenGenes_RESULTS_dada2_asv.fa
## 792
## make_12s_16s_simulated_reads_6-fakeGenes_Random_RESULTS_dada2_asv.fa
## 800
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_12S_RESULTS_dada2_asv.fa
## 259
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_16S_RESULTS_dada2_asv.fa
## 287
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa
## 243
## make_12s_16s_simulated_reads_8-Rottnest_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa
## 1053
## make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_12S_RESULTS_dada2_asv.fa
## 1070
## make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_16S_RESULTS_dada2_asv.fa
## 1253
table(sixteenS_data$Subject)
##
## 16S_v04_final.fasta
## 14732
table(co1_data$Query)
##
## KWest_16S_PooledTrue.fa
## 1641
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa
## 595
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_noLulu_RESULTS_dada2_asv.fa
## 595
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa
## 597
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_noLulu_RESULTS_dada2_asv.fa
## 597
## make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa
## 1102
## make_12s_16s_simulated_reads_6-fakeGenes_GreenGenes_RESULTS_dada2_asv.fa
## 594
## make_12s_16s_simulated_reads_6-fakeGenes_Random_RESULTS_dada2_asv.fa
## 600
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_12S_RESULTS_dada2_asv.fa
## 144
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_16S_RESULTS_dada2_asv.fa
## 162
## make_12s_16s_simulated_reads_7-Lutjanids_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa
## 300
## make_12s_16s_simulated_reads_8-Rottnest_Mock_runEDNAFlow_CO1_RESULTS_dada2_asv.fa
## 1327
## make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_12S_RESULTS_dada2_asv.fa
## 613
## make_12s_16s_simulated_reads_8-Rottnest_runEDNAFLOW_16S_RESULTS_dada2_asv.fa
## 674
twelveS_data_vs_12S_100 <- twelveS_data |> filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa')
sixteenS_data_vs_16S_100 <- sixteenS_data |> filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa' )
co1_data_vs_co1_100 <- co1_data |> filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa')
twelveS_data_vs_12S_100 |> select(Type, species) |> filter(species != 'dropped' &
!is.na(species)) |>
group_by(Type) |> count(species) |> summarise(n = n()) |>
ggplot(aes(x = Type, y = n, fill = n)) + geom_col() + coord_flip() +
theme_minimal() +
ylab('Count') +
ggtitle('12S: Species-level hits per classifier')
twelveS_data_vs_12S_100 |> select(Type, genus) |> filter(genus != 'dropped' &
!is.na(genus)) |>
group_by(Type) |> count(genus) |> summarise(n = n()) |>
ggplot(aes(x = Type, y = n, fill = n)) + geom_col() + coord_flip() +
theme_minimal() +
ylab('Count') +
ggtitle('12S: Genus-level hits per classifier')
twelveS_truth <- truth |> filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa') |> select(OTU, family, species) |> rename(True_OTU = OTU, True_family = family, True_species = species)
head(twelveS_truth)
## # A tibble: 6 Ă— 3
## True_OTU True_family True_species
## <chr> <chr> <chr>
## 1 ASV_1 Syngnathidae Phyllopteryx taeniolatus
## 2 ASV_2 Carcharhinidae Glyphis garricki
## 3 ASV_3 Mullidae Parupeneus barberinus
## 4 ASV_4 Holocentridae Myripristis vittata
## 5 ASV_5 Scincidae Tropidophorus hainanus
## 6 ASV_6 Anatidae Aythya nyroca
twelveS_data_vs_12S_100 |> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
mutate(Correct = True_species == species) |>
filter(species != 'dropped' & !is.na(species)) |>
group_by(Type) |> count(Correct) |>
ggplot(aes(x = fct_rev(fct_reorder2(Type, Correct, n)), fill = Correct, y = n))+ geom_col() +
coord_flip() + theme_minimal() + xlab('Type') +
ggtitle('12S: Correct and incorrect species-level classifications (absolute)') +
scale_fill_manual(values = c("#E69F00", "#56B4E9", "#009E73",
"#F0E442", "#0072B2", "#D55E00", "#CC79A7"))
cols <- c('Correct species' = "#009E73", 'Correct genus'="#56B4E9", 'Correct family' = "#0072B2", 'Incorrect family' = "#E69F00", 'Incorrect genus'="#F0E442", 'Incorrect species'="#D55E00", 'No hit'= "#CC79A7")
twelve_s_relative_plot <- twelveS_data_vs_12S_100 |> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type) |>
count(CorrectSpecies) |>
mutate(total = sum(n, na.rm=TRUE)) |>
mutate(missing = 99 - total) |>
group_modify(~ add_row(.x)) |>
group_modify(~ {
.x |> mutate(new_col= max(missing, na.rm=TRUE)) |>
mutate(n = case_when(is.na(CorrectSpecies) & is.na(missing) ~ new_col,
TRUE ~ n)) |>
select(-new_col)
} ) |>
mutate(total = 99) |>
mutate(perc = n / total * 100) |>
mutate(CorrectSpecies = replace_na(CorrectSpecies, 'No hit')) |>
mutate(CorrectSpecies = factor(CorrectSpecies, rev(c('Correct species', 'Correct genus', 'Correct family', 'Incorrect family', 'Incorrect genus', 'Incorrect species', 'No hit')))) |>
ggplot(aes(x = fct_rev(fct_reorder2(Type, CorrectSpecies, n)), fill = CorrectSpecies, y = perc))+
geom_col() +
coord_flip() +
theme_minimal() +
ylab('Percentage') + xlab('Type') +
ggtitle('12S: Correct and incorrect species-level classifications (relative)') +
scale_fill_manual(name = 'Outcome', values = cols, breaks=names(cols))
twelve_s_relative_plot
### Calculate Upset-based species sightings
type_list <- twelveS_data_vs_12S_100 |> select(Type, species) |> unique() |> filter(!is.na(species) & species != 'dropped') |>
group_by(species) |>
summarize('Type' = list(Type))
a <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('12S: Shared species') +
ylab('Species')
a
type_list <- twelveS_data_vs_12S_100 |> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
filter(species == True_species) |>
filter(species != 'dropped' & !is.na(species)) |>
select(Type, species) |> unique() |>
group_by(species) |>
summarize('Type' = list(Type))
b <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('12S: Shared correct species') +
ylab('Species')
b
type_list <- twelveS_data_vs_12S_100 |> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
filter(species != True_species) |>
filter(species != 'dropped' & !is.na(species)) |>
select(Type, species) |> unique() |>
group_by(species) |>
summarize('Type' = list(Type))
c <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('12S: Shared incorrect species') +
ylab('Species')
c
a + b + c & ylim(c(0, 30)) &
theme(
# Hide panel borders and remove grid lines
panel.border = element_blank(),
panel.grid.major = element_blank(),
panel.background = element_blank(),
panel.grid.minor = element_blank(),
#panel.grid.major.y = element_line(),
# Change axis line
axis.line = element_line(colour = "black")
)
add_scores <- function(twelveS_data_vs_12S_100_with_MaxTruth, twelveS_truth ) {
twelveS_data_vs_12S_100_with_MaxTruth|> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums))
}
scores <- add_scores(twelveS_data_vs_12S_100, twelveS_truth)
scores <- scores |> rowwise() |> mutate(Recall = recall(TP, FN), Precision = precision(TP, FP),
f1 = f1(Precision, Recall), f0.5 = f0.5(Precision, Recall), accuracy = accuracy(TP, FP, FN, TN))
scores
## # A tibble: 12 Ă— 10
## # Rowwise:
## Type TP FP TN FN Recall Precision f1 f0.5 accuracy
## <chr> <int> <int> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 BLAST100 59 6 0 34 0.634 0.908 0.747 0.836 0.596
## 2 BLAST97 48 8 0 43 0.527 0.857 0.653 0.762 0.485
## 3 CustomNBC 42 19 0 38 0.525 0.689 0.596 0.648 0.424
## 4 Kraken_0.0 54 14 0 31 0.635 0.794 0.706 0.756 0.545
## 5 Kraken_0.05 51 7 0 41 0.554 0.879 0.68 0.787 0.515
## 6 Kraken_0.1 47 5 0 47 0.5 0.904 0.644 0.778 0.475
## 7 MMSeqs2_100 59 6 0 34 0.634 0.908 0.747 0.836 0.596
## 8 MMSeqs2_97 59 13 0 27 0.686 0.819 0.747 0.789 0.596
## 9 Metabuli 57 16 0 26 0.687 0.781 0.731 0.76 0.576
## 10 Mothur 41 20 0 38 0.519 0.672 0.586 0.635 0.414
## 11 Qiime2 29 22 0 48 0.377 0.569 0.453 0.516 0.293
## 12 VSEARCH 40 15 0 44 0.476 0.727 0.576 0.658 0.404
twelveS_scoreS_plot <- scores |> select(-c(TP, FP, FN, TN)) |> pivot_longer(-Type, names_to='Score') |> ggplot(aes(x = fct_rev(fct_reorder(Type, value)), y = value, group=Score, color = Score, fill =Score)) + geom_line() + ylim(c(0, 1.05)) + theme_minimal_hgrid()+ theme(axis.text.x = element_text( angle = 45, hjust = 1)) + ylab('Score') + xlab('Tool') + ggtitle('12S scores')
twelveS_scoreS_plot
scores |> select(-c(TP, FP, TN, FN)) |>
rename('group' = 'Type') |>
ggradar()
Let’s also make a heatmap from that
b <- scores$Type
m <- scores |> select(-Type) |> select(accuracy, Recall, Precision, f1, f0.5) |> as.matrix()
rownames(m) <- b
pheatmap(m, cluster_cols=FALSE, display_numbers = TRUE,
color = colorRampPalette(brewer.pal(n = 7, name =
"RdYlGn"))(100))
b <- scores$Type
m <- scores |> select(-Type) |> select(TP, FP, FN) |> as.matrix()
rownames(m) <- b
pheatmap(m, cluster_cols=FALSE, display_numbers = TRUE,
color = colorRampPalette(brewer.pal(n = 7, name =
"RdYlGn"))(100))
table(twelveS_data_vs_12S_100$Type)
##
## BLAST100 BLAST97 CustomNBC Kraken_0.0 Kraken_0.05 Kraken_0.1
## 86 95 99 99 99 99
## Metabuli MMSeqs2_100 MMSeqs2_97 Mothur Qiime2 VSEARCH
## 99 99 99 99 51 99
First, we count the per-OTU species hits
twelveS_data_vs_12S_100_maxCount <- twelveS_data_vs_12S_100 |>
mutate(species = na_if(species, 'dropped')) |>
filter(!is.na(species)) |>
#filter(! Type %in% c('Mothur', 'VSEARCH', 'Kraken_0.2', 'Qiime2', 'Metabuli', 'NBC', 'BLAST97', 'Kraken_0.0', 'Kraken_0.1')) |>
group_by(Query, Subject, OTU) |>
count(species) |>
# double check the truth
#left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
#mutate(Truth = True_species == species) |>
# pull out the per-group highest n
filter( n > 4) |>
slice_max(n, n=1, with_ties = FALSE) |>
mutate(Type = 'MaxCount', .before = 'Query') |>
select(-n)
twelveS_data_vs_12S_100_maxCount
## # A tibble: 67 Ă— 5
## # Groups: Query, Subject, OTU [67]
## Type Query Subject OTU species
## <chr> <chr> <chr> <chr> <chr>
## 1 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_1 Phyllo…
## 2 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Cirrip…
## 3 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Sterco…
## 4 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Carcha…
## 5 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Hemigy…
## 6 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Ctenoc…
## 7 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Daptio…
## 8 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Engrau…
## 9 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Bathyr…
## 10 MaxCount make_12s_16s_simulated_reads_5-BetterDatabase… 12s_v0… ASV_… Tripho…
## # ℹ 57 more rows
twelveS_data_vs_12S_100_with_MaxTruth <- twelveS_data_vs_12S_100 |>
bind_rows(twelveS_data_vs_12S_100_maxCount) #|>
#filter(! Type %in% c('Mothur', 'VSEARCH', 'Kraken_0.2', 'Qiime2', 'Metabuli', 'NBC', 'BLAST97', 'Kraken_0.0', 'Kraken_0.1'))
maxTruth_scores <- add_scores(twelveS_data_vs_12S_100_with_MaxTruth, twelveS_truth )
maxTruth_scores <- maxTruth_scores |> rowwise() |> mutate(Recall = recall(TP, FN), Precision = precision(TP, FP),
f1 = f1(Precision, Recall), f0.5 = f0.5(Precision, Recall), accuracy = accuracy(TP, FP, FN, TN))
maxTruth_scoreS_plot <- maxTruth_scores |> select(-c(TP, FP, FN, TN)) |> pivot_longer(-Type, names_to='Score') |> ggplot(aes(x = fct_rev(fct_reorder(Type, value)), y = value, group=Score, color = Score, fill =Score)) + geom_line() + ylim(c(0, 1)) + theme_minimal_hgrid()+ theme(axis.text.x = element_text( angle = 45, hjust = 1)) + ylab('Score') + xlab('Tool') + geom_point() + ggtitle('12S scores')
maxTruth_scoreS_plot
Interestingly, just counting the labels is not good! It performs worse
than BLAST.
sixteenS_truth <- truth |> filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa') |> select(OTU, family, species) |> rename(True_OTU = OTU, True_family = family, True_species = species)
head(sixteenS_truth)
## # A tibble: 6 Ă— 3
## True_OTU True_family True_species
## <chr> <chr> <chr>
## 1 ASV_1 Syngnathidae Phyllopteryx taeniolatus
## 2 ASV_2 Carcharhinidae Glyphis garricki
## 3 ASV_3 Merlucciidae Merluccius productus
## 4 ASV_4 Mullidae Parupeneus barberinus
## 5 ASV_5 Syngnathidae Hippocampus algiricus
## 6 ASV_6 Eleotridae Bostrychus sinensis
sixteenS_relative_plot <- sixteenS_data_vs_16S_100 |> left_join(sixteenS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type) |>
count(CorrectSpecies) |>
mutate(total = sum(n)) |>
mutate(missing = 99 - total) |>
group_modify(~ add_row(.x)) |>
group_modify(~ {
.x |> mutate(new_col= max(missing, na.rm=TRUE)) |>
mutate(n = case_when(is.na(CorrectSpecies) & is.na(missing) ~ new_col,
TRUE ~ n)) |>
select(-new_col)
} ) |>
mutate(total = 99) |>
mutate(perc = n / total * 100) |>
mutate(CorrectSpecies = replace_na(CorrectSpecies, 'No hit')) |>
mutate(CorrectSpecies = factor(CorrectSpecies, rev(c('Correct species', 'Correct genus', 'Correct family', 'Incorrect family', 'Incorrect genus', 'Incorrect species', 'No hit')))) |>
tidyr::complete(CorrectSpecies, fill = list(n=0, total = 99, missing = NA, perc = 0)) |>
ggplot(aes(x = fct_rev(fct_reorder2(Type, CorrectSpecies, n)), fill = CorrectSpecies, y = perc))+
geom_col() +
coord_flip() +
theme_minimal() +
ylab('Percentage') + xlab('Type') +
ggtitle('16S: Correct and incorrect species-level classifications (relative)') +
scale_fill_manual(name = 'Outcome', values = cols, breaks=names(cols))
sixteenS_relative_plot
scores <- add_scores(sixteenS_data_vs_16S_100, sixteenS_truth)
scores
## # A tibble: 12 Ă— 5
## Type TP FP TN FN
## <chr> <int> <int> <int> <dbl>
## 1 BLAST100 51 0 0 48
## 2 BLAST97 48 5 0 46
## 3 CustomNBC 52 10 0 37
## 4 Kraken_0.0 52 15 0 32
## 5 Kraken_0.05 48 13 0 38
## 6 Kraken_0.1 45 10 0 44
## 7 MMSeqs2_100 51 0 0 48
## 8 MMSeqs2_97 60 5 0 34
## 9 Metabuli 55 21 0 23
## 10 Mothur 50 14 0 35
## 11 Qiime2 42 15 0 42
## 12 VSEARCH 43 12 0 44
scores <- scores |> rowwise() |> mutate(Recall = recall(TP, FN), Precision = precision(TP, FP),
f1 = f1(Precision, Recall), f0.5 = f0.5(Precision, Recall), accuracy = accuracy(TP, FP, FN, TN))
scores
## # A tibble: 12 Ă— 10
## # Rowwise:
## Type TP FP TN FN Recall Precision f1 f0.5 accuracy
## <chr> <int> <int> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 BLAST100 51 0 0 48 0.515 1 0.68 0.842 0.515
## 2 BLAST97 48 5 0 46 0.511 0.906 0.653 0.784 0.485
## 3 CustomNBC 52 10 0 37 0.584 0.839 0.689 0.772 0.525
## 4 Kraken_0.0 52 15 0 32 0.619 0.776 0.689 0.739 0.525
## 5 Kraken_0.05 48 13 0 38 0.558 0.787 0.653 0.727 0.485
## 6 Kraken_0.1 45 10 0 44 0.506 0.818 0.625 0.728 0.455
## 7 MMSeqs2_100 51 0 0 48 0.515 1 0.68 0.842 0.515
## 8 MMSeqs2_97 60 5 0 34 0.638 0.923 0.755 0.847 0.606
## 9 Metabuli 55 21 0 23 0.705 0.724 0.714 0.720 0.556
## 10 Mothur 50 14 0 35 0.588 0.781 0.671 0.733 0.505
## 11 Qiime2 42 15 0 42 0.5 0.737 0.596 0.673 0.424
## 12 VSEARCH 43 12 0 44 0.494 0.782 0.606 0.700 0.434
sixteenS_score_plot <- scores |> select(-c(TP, FP, FN, TN)) |> pivot_longer(-Type, names_to='Score') |> ggplot(aes(x = fct_rev(fct_reorder(Type, value)), y = value, group=Score, color = Score, fill =Score)) + geom_line() + ylim(c(0, 1)) + theme_minimal_hgrid() + theme(axis.text.x = element_text( angle = 45, hjust = 1)) + ylab('Score') + xlab('Tool') + ggtitle('16S scores')
sixteenS_score_plot
b <- scores$Type
m <- scores |> select(-Type) |> select(accuracy, Recall, Precision, f1, f0.5) |> as.matrix()
rownames(m) <- b
pheatmap(m, cluster_cols=FALSE, display_numbers = TRUE,
color = colorRampPalette(brewer.pal(n = 7, name =
"RdYlGn"))(100))
b <- scores$Type
m <- scores |> select(-Type) |> select(TP, FP, FN) |> as.matrix()
rownames(m) <- b
pheatmap(m, cluster_cols=FALSE, display_numbers = TRUE,
color = colorRampPalette(brewer.pal(n = 7, name =
"RdYlGn"))(100))
### Calculate Upset-based species sightings
type_list <- sixteenS_data_vs_16S_100 |> select(Type, species) |> unique() |> filter(!is.na(species) & species != 'dropped') |>
group_by(species) |>
summarize('Type' = list(Type))
a <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('16S: Shared species') +
ylab('Species')
a
type_list <- sixteenS_data_vs_16S_100 |> left_join(sixteenS_truth, by = c('OTU' = 'True_OTU')) |>
filter(species == True_species) |>
filter(species != 'dropped' & !is.na(species)) |>
select(Type, species) |> unique() |>
group_by(species) |>
summarize('Type' = list(Type))
b <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('16S: Shared correct species') +
ylab('Species')
b
type_list <- sixteenS_data_vs_16S_100 |> left_join(sixteenS_truth, by = c('OTU' = 'True_OTU')) |>
filter(species != True_species) |>
filter(species != 'dropped' & !is.na(species)) |>
select(Type, species) |> unique() |>
group_by(species) |>
summarize('Type' = list(Type))
c <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('16S: Shared incorrect species') +
ylab('Species')
c
a + b + c & ylim(c(0, 20)) &
theme(
# Hide panel borders and remove grid lines
panel.border = element_blank(),
panel.grid.major = element_blank(),
panel.background = element_blank(),
panel.grid.minor = element_blank(),
#panel.grid.major.y = element_line(),
# Change axis line
axis.line = element_line(colour = "black")
)
CO1_truth <- truth |> filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa') |> select(OTU, family, species) |> rename(True_OTU = OTU, True_family = family, True_species = species)
head(CO1_truth)
## # A tibble: 6 Ă— 3
## True_OTU True_family True_species
## <chr> <chr> <chr>
## 1 ASV_1 Syngnathidae Phycodurus eques
## 2 ASV_2 Syngnathidae Phyllopteryx taeniolatus
## 3 ASV_3 Anatidae dropped
## 4 ASV_4 Carcharhinidae Glyphis garricki
## 5 ASV_5 Syngnathidae dropped
## 6 ASV_6 Eleotridae Bostrychus sinensis
CO1_relative_plot <- co1_data_vs_co1_100 |> left_join(CO1_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type) |>
count(CorrectSpecies) |>
mutate(total = sum(n)) |>
mutate(missing = 99 - total) |>
group_modify(~ add_row(.x)) |>
group_modify(~ {
.x |> mutate(new_col= max(missing, na.rm=TRUE)) |>
mutate(n = case_when(is.na(CorrectSpecies) & is.na(missing) ~ new_col,
TRUE ~ n)) |>
select(-new_col)
} ) |>
mutate(total = 99) |>
mutate(perc = n / total * 100) |>
mutate(CorrectSpecies = replace_na(CorrectSpecies, 'No hit')) |>
mutate(CorrectSpecies = factor(CorrectSpecies, rev(c('Correct species', 'Correct genus', 'Correct family', 'Incorrect family', 'Incorrect genus', 'Incorrect species', 'No hit')))) |>
tidyr::complete(CorrectSpecies, fill = list(n=0, total = 99, missing = NA, perc = 0)) |>
ggplot(aes(x = fct_rev(fct_reorder2(Type, CorrectSpecies, n)), fill = CorrectSpecies, y = perc))+
geom_col() +
coord_flip() +
theme_minimal() +
ylab('Percentage') + xlab('Type') +
ggtitle('CO1: Correct and incorrect species-level classifications (relative)') +
scale_fill_manual(name = 'Outcome', values = cols, breaks=names(cols))
CO1_relative_plot
scores <- add_scores(co1_data_vs_co1_100, CO1_truth)
scores
## # A tibble: 12 Ă— 5
## Type TP FP TN FN
## <chr> <int> <int> <int> <dbl>
## 1 BLAST100 48 5 0 46
## 2 BLAST97 46 6 0 47
## 3 CustomNBC 51 24 0 24
## 4 Kraken_0.0 48 6 0 45
## 5 Kraken_0.05 38 2 0 59
## 6 Kraken_0.1 27 2 0 70
## 7 MMSeqs2_100 0 0 0 99
## 8 MMSeqs2_97 0 0 0 99
## 9 Metabuli 33 14 0 52
## 10 Mothur 52 27 0 20
## 11 Qiime2 49 15 0 35
## 12 VSEARCH 0 0 0 99
scores <- scores |> rowwise() |> mutate(Recall = recall(TP, FN), Precision = precision(TP, FP),
f1 = f1(Precision, Recall), f0.5 = f0.5(Precision, Recall), accuracy = accuracy(TP, FP, FN, TN))
scores
## # A tibble: 12 Ă— 10
## # Rowwise:
## Type TP FP TN FN Recall Precision f1 f0.5 accuracy
## <chr> <int> <int> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 BLAST100 48 5 0 46 0.511 0.906 0.653 0.784 0.485
## 2 BLAST97 46 6 0 47 0.495 0.885 0.634 0.764 0.465
## 3 CustomNBC 51 24 0 24 0.68 0.68 0.68 0.68 0.515
## 4 Kraken_0.0 48 6 0 45 0.516 0.889 0.653 0.777 0.485
## 5 Kraken_0.05 38 2 0 59 0.392 0.95 0.555 0.739 0.384
## 6 Kraken_0.1 27 2 0 70 0.278 0.931 0.429 0.634 0.273
## 7 MMSeqs2_100 0 0 0 99 0 0 0 0 0
## 8 MMSeqs2_97 0 0 0 99 0 0 0 0 0
## 9 Metabuli 33 14 0 52 0.388 0.702 0.5 0.604 0.333
## 10 Mothur 52 27 0 20 0.722 0.658 0.689 0.670 0.525
## 11 Qiime2 49 15 0 35 0.583 0.766 0.662 0.721 0.495
## 12 VSEARCH 0 0 0 99 0 0 0 0 0
CO1_score_plot <- scores |> select(-c(TP, FP, FN, TN)) |> pivot_longer(-Type, names_to='Score') |> ggplot(aes(x = fct_rev(fct_reorder(Type, value)), y = value, group=Score, color = Score, fill =Score)) + geom_line() + ylim(c(0, 1)) + theme_minimal_hgrid() + theme(axis.text.x = element_text( angle = 45, hjust = 1)) + ylab('Score') + xlab('Tool') + ggtitle('CO1 scores')
CO1_score_plot
b <- scores$Type
m <- scores |> select(-Type) |> select(accuracy, Recall, Precision, f1, f0.5) |> as.matrix()
rownames(m) <- b
pheatmap(m, cluster_cols=FALSE, display_numbers = TRUE,
color = colorRampPalette(brewer.pal(n = 7, name =
"RdYlGn"))(100))
b <- scores$Type
m <- scores |> select(-Type) |> select(TP, FP, FN) |> as.matrix()
rownames(m) <- b
pheatmap(m, cluster_cols=FALSE, display_numbers = TRUE,
color = colorRampPalette(brewer.pal(n = 7, name =
"RdYlGn"))(100))
### Calculate Upset-based species sightings
type_list <- co1_data_vs_co1_100 |> select(Type, species) |> unique() |> filter(!is.na(species) & species != 'dropped') |>
group_by(species) |>
summarize('Type' = list(Type))
a <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('CO1: Shared species') +
ylab('Species')
a
type_list <- co1_data_vs_co1_100 |> left_join(CO1_truth, by = c('OTU' = 'True_OTU')) |>
filter(species == True_species) |>
filter(species != 'dropped' & !is.na(species)) |>
select(Type, species) |> unique() |>
group_by(species) |>
summarize('Type' = list(Type))
b <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('CO1: Shared correct species') +
ylab('Species')
b
type_list <- co1_data_vs_co1_100 |> left_join(CO1_truth, by = c('OTU' = 'True_OTU')) |>
filter(species != True_species) |>
filter(species != 'dropped' & !is.na(species)) |>
select(Type, species) |> unique() |>
group_by(species) |>
summarize('Type' = list(Type))
c <- type_list |>
ggplot(aes(x = Type)) +
geom_bar() +
scale_x_upset(n_intersections = 12) +
ggtitle('CO1: Shared incorrect species') +
ylab('Species')
c
a + b + c & ylim(c(0, 20)) &
theme(
# Hide panel borders and remove grid lines
panel.border = element_blank(),
panel.grid.major = element_blank(),
panel.background = element_blank(),
panel.grid.minor = element_blank(),
#panel.grid.major.y = element_line(),
# Change axis line
axis.line = element_line(colour = "black")
)
sixteenS_relative_plot / twelve_s_relative_plot / CO1_relative_plot
Let’s make without titles, but with a/b
(sixteenS_relative_plot + ggtitle('') + ylab(''))/ (twelve_s_relative_plot + ggtitle('')) / (CO1_relative_plot + ggtitle('')) +
plot_annotation(tag_levels = c('A','B')) +
plot_layout(guides = 'collect')
(sixteenS_score_plot +geom_point() + theme(axis.title.x = element_blank()))/ (twelveS_scoreS_plot + geom_point()) / (CO1_score_plot + geom_point())
twelve_exclusions <- data |> filter(str_starts(Subject, '12s_v010_final.fasta_Taxonomies.CountedFams.txt_')) |>
filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_12S_Lulu_RESULTS_dada2_asv.fa')
twelve_exclusions_split <- twelve_exclusions |> separate(Subject, into = c('before', 'hit'), sep='.txt_') |>
# get rid of leftover non-subsetted databases
filter(!is.na(hit)) |>
separate(hit, into=c('Database', 'after'), sep='_subset')
twelve_exclusions_split_averaged <- twelve_exclusions_split |> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type, Database, after) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums)) |>
group_by(Type, Database) |>
summarise(mean_TP = mean(TP),
mean_FP = mean(FP),
mean_TN = mean(TN),
mean_FN = mean(FN)) |>
rowwise() |>
mutate(Recall = recall(mean_TP, mean_FN),
Precision = precision(mean_TP, mean_FP),
f1 = f1(Precision, Recall),
f0.5 = f0.5(Precision, Recall),
accuracy = accuracy(mean_TP, mean_FP, mean_FN, mean_TN))
## `summarise()` has grouped output by 'Type', 'Database'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'Type'. You can override using the
## `.groups` argument.
twelve_exclusions_split_averaged <- twelve_exclusions_split_averaged |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%', # we switch from seventy INCLUDED to 30 EXCLUDED
Database == 'thirty' ~ '70%',
TRUE ~ Database))
f1_pl <- twelve_exclusions_split_averaged |>
ggplot(aes(x = Type, y = f1, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
f0.5_pl <- twelve_exclusions_split_averaged |>
ggplot(aes(x = Type, y = f0.5, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
Precision_pl <- twelve_exclusions_split_averaged |>
ggplot(aes(x = Type, y = Precision, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
Recall_pl <- twelve_exclusions_split_averaged |>
ggplot(aes(x = Type, y = Recall, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
(f1_pl / f0.5_pl / Precision_pl / Recall_pl) + plot_layout(guides = 'collect')
Lets zero in on the Precision and make boxplots with jitter dots
un_summarised_twelve <- twelve_exclusions_split |> left_join(twelveS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type, Database, after) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums)) |>
rowwise() |>
mutate(Recall = recall(TP, FN),
Precision = precision(TP, FP),
f1 = f1(Precision, Recall),
f0.5 = f0.5(Precision, Recall),
accuracy = accuracy(TP, FP, FN, TN)) |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database))
## `summarise()` has grouped output by 'Type', 'Database'. You can override using
## the `.groups` argument.
un_summarised_twelve |> group_by(Type, Database) |> mutate(best = max(mean(Precision))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = Precision)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('Precision') +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_twelve |> group_by(Type, Database) |> mutate(best = max(mean(f0.5))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = f0.5)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('f0.5') +
ylim(c(0, 1)) +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_twelve |> group_by(Type, Database) |> mutate(best = max(mean(Recall))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = Recall)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('f0.5') +
ylim(c(0, 1)) +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_twelve |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Qiime2', 'TNT')) |>
ggplot(aes(x=Database, y = Precision, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot() +
labs(fill='Type') +
ylab('Precision') +
theme_minimal()
false_positives <- un_summarised_twelve |>
filter(Type %in% c('BLAST100', 'Kraken_0.0','Metabuli', 'Kraken_0.1', 'MMSeqs2', 'TNT')) |>
ggplot(aes(x=Database, y = FP/99*100, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot(outlier.shape=NA) +
labs(fill='Type') +
ylab('False positives (%)') +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE)+
theme_minimal()
false_positives
true_positives <- un_summarised_twelve |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1', 'MMSeqs2', 'TNT')) |>
ggplot(aes(x=Database, y = TP/99*100, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot(outlier.shape=NA) +
labs(fill='Type') +
ylab('True positives (%)') +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE)+
theme_minimal()
true_positives
false_positives/ true_positives + plot_layout(guides = 'collect') & coord_flip()
### Phylogenetic diversity
We can also easily calculate alpha diversity across these tools, as alpha diversity is the number of species. We treat classifiers/Types as sites.
spec_summarised <- twelve_exclusions_split |>
group_by(Type, Query, Database, after) |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database)) |>
filter(!is.na(species)) |>
summarise(`Alpha diversity index` = length(unique(species)))
## `summarise()` has grouped output by 'Type', 'Query', 'Database'. You can
## override using the `.groups` argument.
spec_summarised |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot() +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) + coord_flip() + theme_minimal()
Let’s also do not all of the classifiers
spec_summarised |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot() +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) + coord_flip() + theme_minimal()
a <- spec_summarised |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0','Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
group_by(Database) |>
arrange(Database) |>
group_map(~aov(`Alpha diversity index` ~ Type, data=.))
names(a) <- spec_summarised |> arrange(Database) |> pull(Database) |> unique() # I don't like this :(
a
## $`30%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 8875.2 1211.2
## Deg. of Freedom 5 54
##
## Residual standard error: 4.735993
## Estimated effects may be unbalanced
##
## $`50%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 12781.2 1197.8
## Deg. of Freedom 5 54
##
## Residual standard error: 4.709722
## Estimated effects may be unbalanced
##
## $`70%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 17126.53 703.40
## Deg. of Freedom 5 54
##
## Residual standard error: 3.609145
## Estimated effects may be unbalanced
groupslist <- list()
for(key in names(a)) {
print(key)
groupslist[[key]] <- HSD.test(a[[key]], 'Type')$groups|>
as_tibble(rownames = 'Type') |>
select(-`Alpha diversity index`)
}
## [1] "30%"
## [1] "50%"
## [1] "70%"
groups_df <- bind_rows(groupslist, .id='Database')
twelve_s_exclusions_fig <- spec_summarised |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0','Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
left_join(groups_df, by = c('Database', 'Type')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot(outlier.shape=NA) +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) +
geom_text(aes(x = Type, y = max(`Alpha diversity index`) + 2, label = groups),
#col = 'black',
size = 4) +
#coord_flip() +
theme_minimal() +
theme(axis.text.x = element_text( angle = 90, hjust = 1)) +
guides(fill="none")
twelve_s_exclusions_fig
twelve_spec_summarised <- spec_summarised
twelve_spec_summarised$gene <- '12S'
sixteen_exclusions <- data |> filter(str_starts(Subject, '16S_v04_final.fasta_Taxonomies.')) |>
filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_16S_Lulu_RESULTS_dada2_asv.fa')
table(sixteen_exclusions$Subject)
##
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 1018
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 954
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 980
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 1024
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 1028
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 999
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 981
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 983
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 965
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 994
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 988
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 1045
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 1057
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 1040
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 1047
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 1065
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 1041
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 1047
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 1051
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 1050
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 924
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 964
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 943
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 929
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 960
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 930
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 928
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 951
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 930
## 16S_v04_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 946
sixteen_exclusions_split <- sixteen_exclusions |> separate(Subject, into = c('before', 'hit'), sep='.txt_') |>
# get rid of leftover non-subsetted databases
filter(!is.na(hit)) |>
separate(hit, into=c('Database', 'after'), sep='_subset')
sixteen_exclusions_split_averaged <- sixteen_exclusions_split |> left_join(sixteenS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type, Database, after) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums)) |>
group_by(Type, Database) |>
summarise(mean_TP = mean(TP),
mean_FP = mean(FP),
mean_TN = mean(TN),
mean_FN = mean(FN)) |>
rowwise() |>
mutate(Recall = recall(mean_TP, mean_FN),
Precision = precision(mean_TP, mean_FP),
f1 = f1(Precision, Recall),
f0.5 = f0.5(Precision, Recall),
accuracy = accuracy(mean_TP, mean_FP, mean_FN, mean_TN))
## `summarise()` has grouped output by 'Type', 'Database'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'Type'. You can override using the
## `.groups` argument.
sixteen_exclusions_split_averaged <- sixteen_exclusions_split_averaged |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database))
f1_pl <- sixteen_exclusions_split_averaged |>
ggplot(aes(x = Type, y = f1, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
f0.5_pl <- sixteen_exclusions_split_averaged |>
ggplot(aes(x = Type, y = f0.5, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
Precision_pl <- sixteen_exclusions_split_averaged |>
ggplot(aes(x = Type, y = Precision, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
Recall_pl <- sixteen_exclusions_split_averaged |>
ggplot(aes(x = Type, y = Recall, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
(f1_pl / f0.5_pl / Precision_pl / Recall_pl) + plot_layout(guides = 'collect')
Lets zero in on the Precision and make boxplots with jitter dots
un_summarised_sixteen <- sixteen_exclusions_split |> left_join(sixteenS_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type, Database, after) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums)) |>
rowwise() |>
mutate(Recall = recall(TP, FN),
Precision = precision(TP, FP),
f1 = f1(Precision, Recall),
f0.5 = f0.5(Precision, Recall),
accuracy = accuracy(TP, FP, FN, TN)) |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database))
## `summarise()` has grouped output by 'Type', 'Database'. You can override using
## the `.groups` argument.
un_summarised_sixteen |> group_by(Type, Database) |> mutate(best = max(mean(Precision, na.rm=TRUE))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = Precision)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('Precision') +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_sixteen |> group_by(Type, Database) |> mutate(best = max(mean(f0.5, na.rm=TRUE))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = f0.5)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('f0.5') +
ylim(c(0, 1)) +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_sixteen |> group_by(Type, Database) |> mutate(best = max(mean(Recall))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = Recall)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('f0.5') +
ylim(c(0, 1)) +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_sixteen |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Qiime2')) |>
ggplot(aes(x=Database, y = Precision, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot() +
labs(fill='Type') +
ylab('Precision') +
theme_minimal()
false_positives <- un_summarised_sixteen |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1', 'MMSeqs2', 'TNT')) |>
ggplot(aes(x=Database, y = FP/99*100, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot(outlier.shape=NA) +
labs(fill='Type') +
ylab('False positives (%)') +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE)+
theme_minimal()
false_positives
true_positives <- un_summarised_sixteen |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1', 'MMSeqs2', 'TNT')) |>
ggplot(aes(x=Database, y = TP/99*100, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot(outlier.shape=NA) +
labs(fill='Type') +
ylab('True positives (%)') +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE)+
theme_minimal()
true_positives
false_positives/ true_positives + plot_layout(guides = 'collect') & coord_flip()
### Phylogenetic diversity
We can also easily calculate alpha diversity across these tools, as alpha diversity is the number of species. We treat classifiers/Types as sites.
spec_summarised <- sixteen_exclusions_split |>
group_by(Type, Query, Database, after) |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database)) |>
filter(!is.na(species)) |>
summarise(`Alpha diversity index` = length(unique(species)))
## `summarise()` has grouped output by 'Type', 'Query', 'Database'. You can
## override using the `.groups` argument.
spec_summarised |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot() +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) + coord_flip() + theme_minimal()
Let’s also do not all of the classifiers
spec_summarised |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot() +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) + coord_flip() + theme_minimal()
a <- spec_summarised |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
group_by(Database) |>
arrange(Database) |>
group_map(~aov(`Alpha diversity index` ~ Type, data=.))
names(a) <- spec_summarised |> arrange(Database) |> pull(Database) |> unique() # I don't like this :(
a
## $`30%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 11466.261 396.722
## Deg. of Freedom 5 53
##
## Residual standard error: 2.735932
## Estimated effects may be unbalanced
##
## $`50%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 17652.28 2570.30
## Deg. of Freedom 5 54
##
## Residual standard error: 6.899141
## Estimated effects may be unbalanced
##
## $`70%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 20925.344 1478.656
## Deg. of Freedom 5 53
##
## Residual standard error: 5.281966
## Estimated effects may be unbalanced
library(agricolae)
groupslist <- list()
for(key in names(a)) {
print(key)
groupslist[[key]] <- HSD.test(a[[key]], 'Type')$groups|>
as_tibble(rownames = 'Type') |>
select(-`Alpha diversity index`)
}
## [1] "30%"
## [1] "50%"
## [1] "70%"
groups_df <- bind_rows(groupslist, .id='Database')
sixteen_s_exclusions_fig <- spec_summarised |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
left_join(groups_df, by = c('Database', 'Type')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot(outlier.shape=NA) +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) +
geom_text(aes(x = Type, y = max(`Alpha diversity index`) + 2, label = groups),
#col = 'black',
size = 4) +
#coord_flip() +
theme_minimal() +
theme(axis.text.x = element_text( angle = 90, hjust = 1)) +
guides(fill="none")
sixteen_s_exclusions_fig
sixteen_spec_summarised <- spec_summarised
sixteen_spec_summarised$gene <- '16S'
co1_exclusions <- data |> filter(str_starts(Subject, 'c01_v03_final.fasta_Taxonomies.')) |>
filter(Query == 'make_12s_16s_simulated_reads_5-BetterDatabaseARTSimulation_runEDNAFLOW_CO1_RESULTS_dada2_asv.fa')
table(co1_exclusions$Subject)
##
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_1.fasta
## 938
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_10.fasta
## 951
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_2.fasta
## 919
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_3.fasta
## 980
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_4.fasta
## 941
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_5.fasta
## 992
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_6.fasta
## 975
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_7.fasta
## 1002
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_8.fasta
## 946
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_fifty_subset_9.fasta
## 966
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_1.fasta
## 1037
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_10.fasta
## 1021
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_2.fasta
## 1018
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_3.fasta
## 1005
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_4.fasta
## 971
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_5.fasta
## 1028
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_6.fasta
## 1007
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_7.fasta
## 1042
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_8.fasta
## 1062
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_seventy_subset_9.fasta
## 1043
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_1.fasta
## 921
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_10.fasta
## 943
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_2.fasta
## 909
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_3.fasta
## 908
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_4.fasta
## 895
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_5.fasta
## 902
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_6.fasta
## 903
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_7.fasta
## 886
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_8.fasta
## 885
## c01_v03_final.fasta_Taxonomies.CountedFams.txt_thirty_subset_9.fasta
## 886
co1_exclusions_split <- co1_exclusions |> separate(Subject, into = c('before', 'hit'), sep='.txt_') |>
# get rid of leftover non-subsetted databases
filter(!is.na(hit)) |>
separate(hit, into=c('Database', 'after'), sep='_subset')
co1_exclusions_split_averaged <- co1_exclusions_split |> left_join(CO1_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type, Database, after) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums)) |>
group_by(Type, Database) |>
summarise(mean_TP = mean(TP),
mean_FP = mean(FP),
mean_TN = mean(TN),
mean_FN = mean(FN)) |>
rowwise() |>
mutate(Recall = recall(mean_TP, mean_FN),
Precision = precision(mean_TP, mean_FP),
f1 = f1(Precision, Recall),
f0.5 = f0.5(Precision, Recall),
accuracy = accuracy(mean_TP, mean_FP, mean_FN, mean_TN))
## `summarise()` has grouped output by 'Type', 'Database'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'Type'. You can override using the
## `.groups` argument.
co1_exclusions_split_averaged <- co1_exclusions_split_averaged |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database))
f1_pl <- co1_exclusions_split_averaged |>
ggplot(aes(x = Type, y = f1, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
f0.5_pl <- co1_exclusions_split_averaged |>
ggplot(aes(x = Type, y = f0.5, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
Precision_pl <- co1_exclusions_split_averaged |>
ggplot(aes(x = Type, y = Precision, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
Recall_pl <- co1_exclusions_split_averaged |>
ggplot(aes(x = Type, y = Recall, group = Database, color = Database, fill = Database)) +
geom_point() +
geom_line() +
theme_minimal()
(f1_pl / f0.5_pl / Precision_pl / Recall_pl) + plot_layout(guides = 'collect')
Lets zero in on the Precision and make boxplots with jitter dots
un_summarised_co1 <- co1_exclusions_split |> left_join(CO1_truth, by = c('OTU' = 'True_OTU')) |>
separate(True_species, into = c('True_Genus', 'True_Epiteth'), remove = FALSE)|>
mutate(species = na_if(species, 'dropped')) |>
mutate(genus = na_if(genus, 'dropped')) |>
mutate(CorrectSpecies = case_when(!is.na(species) & True_species == species ~ 'Correct species',
!is.na(species) & True_species != species ~ 'Incorrect species',
!is.na(genus) & !is.na(True_Genus) & True_Genus == genus ~ 'Correct genus',
!is.na(genus) & !is.na(True_Genus) & True_Genus != genus ~ 'Incorrect genus',
!is.na(family) & True_family == family ~ 'Correct family',
!is.na(family) & True_family != family ~ 'Incorrect family',
TRUE ~ NA)) |>
group_by(Type, Database, after) |>
summarise(TP = sum(str_detect(CorrectSpecies, pattern='Correct species'), na.rm=TRUE),
FP = sum(str_detect(CorrectSpecies, pattern = 'Incorrect species'), na.rm=TRUE),
TN = sum(str_detect(replace_na(CorrectSpecies,'NA'), pattern='NA') & is.na(True_species), na.rm=TRUE),
FN = sum(is.na(species) & !is.na(True_species))) |>
mutate(sums = TP + FP + TN + FN) |>
mutate(missing = 99 - sums) |>
mutate(FN = FN + missing) |>
mutate(sums = TP + FP + TN + FN) |>
select(-c(missing, sums)) |>
rowwise() |>
mutate(Recall = recall(TP, FN),
Precision = precision(TP, FP),
f1 = f1(Precision, Recall),
f0.5 = f0.5(Precision, Recall),
accuracy = accuracy(TP, FP, FN, TN)) |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database))
## `summarise()` has grouped output by 'Type', 'Database'. You can override using
## the `.groups` argument.
un_summarised_co1 |> group_by(Type, Database) |> mutate(best = max(mean(Precision, na.rm=TRUE))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = Precision)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('Precision') +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_co1 |> group_by(Type, Database) |> mutate(best = max(mean(f0.5, na.rm=TRUE))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = f0.5)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('f0.5') +
ylim(c(0, 1)) +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_co1 |> group_by(Type, Database) |> mutate(best = max(mean(Recall))) |>
ggplot(aes(x = fct_reorder(Type, best), group = interaction(Type, Database), color=Database, y = Recall)) +
geom_boxplot(outlier.shape = NA) +
coord_flip() +
theme_minimal() +
xlab('Type') +
ylab('f0.5') +
ylim(c(0, 1)) +
geom_point(position = position_jitterdodge(), alpha=0.5)
un_summarised_co1 |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Qiime2')) |>
ggplot(aes(x=Database, y = Precision, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot() +
labs(fill='Type') +
ylab('Precision') +
theme_minimal()
false_positives <- un_summarised_co1 |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1', 'MMSeqs2', 'TNT')) |>
ggplot(aes(x=Database, y = FP/99*100, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot(outlier.shape=NA) +
labs(fill='Type') +
ylab('False positives (%)') +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE)+
theme_minimal()
false_positives
true_positives <- un_summarised_co1 |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1', 'MMSeqs2', 'TNT')) |>
ggplot(aes(x=Database, y = TP/99*100, fill=Type)) + #fill=factor(Database, levels=c('30%','50%','70%')))) +
geom_boxplot(outlier.shape=NA) +
labs(fill='Type') +
ylab('True positives (%)') +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE)+
theme_minimal()
true_positives
false_positives/ true_positives + plot_layout(guides = 'collect') & coord_flip()
### Phylogenetic diversity
We can also easily calculate alpha diversity across these tools, as alpha diversity is the number of species. We treat classifiers/Types as sites.
spec_summarised <- co1_exclusions_split |>
group_by(Type, Query, Database, after) |>
mutate(Database = case_when ( Database == 'fifty' ~ '50%',
Database == 'seventy' ~ '30%',
Database == 'thirty' ~ '70%',
TRUE ~ Database)) |>
filter(!is.na(species)) |>
summarise(`Alpha diversity index` = length(unique(species)))
## `summarise()` has grouped output by 'Type', 'Query', 'Database'. You can
## override using the `.groups` argument.
spec_summarised |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot() +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) + coord_flip() + theme_minimal()
Let’s also do not all of the classifiers
spec_summarised |>
filter(Type %in% c('BLAST100', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot() +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) + coord_flip() + theme_minimal()
a <- spec_summarised |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
group_by(Database) |>
arrange(Database) |>
group_map(~aov(`Alpha diversity index` ~ Type, data=.))
names(a) <- spec_summarised |> arrange(Database) |> pull(Database) |> unique() # I don't like this :(
a
## $`30%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 5590.256 758.456
## Deg. of Freedom 5 53
##
## Residual standard error: 3.78292
## Estimated effects may be unbalanced
##
## $`50%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 9280.189 3219.531
## Deg. of Freedom 5 51
##
## Residual standard error: 7.945316
## Estimated effects may be unbalanced
##
## $`70%`
## Call:
## aov(formula = `Alpha diversity index` ~ Type, data = .)
##
## Terms:
## Type Residuals
## Sum of Squares 17609.08 1107.90
## Deg. of Freedom 5 49
##
## Residual standard error: 4.755019
## Estimated effects may be unbalanced
library(agricolae)
groupslist <- list()
for(key in names(a)) {
print(key)
groupslist[[key]] <- HSD.test(a[[key]], 'Type')$groups|>
as_tibble(rownames = 'Type') |>
select(-`Alpha diversity index`)
}
## [1] "30%"
## [1] "50%"
## [1] "70%"
groups_df <- bind_rows(groupslist, .id='Database')
co1_s_exclusions_fig <- spec_summarised |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
left_join(groups_df, by = c('Database', 'Type')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot(outlier.shape=NA) +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) +
geom_text(aes(x = Type, y = max(`Alpha diversity index`) + 2, label = groups),
#col = 'black',
size = 4) +
#coord_flip() +
theme_minimal() +
theme(axis.text.x = element_text( angle = 90, hjust = 1)) +
guides(fill="none")
co1_s_exclusions_fig
co1_spec_summarised <- spec_summarised
co1_spec_summarised$gene <- 'CO1'
both <- rbind(twelve_spec_summarised, sixteen_spec_summarised, co1_spec_summarised)
both_fig <- both |>
filter(Type %in% c('BLAST100', 'BLAST97', 'Kraken_0.0', 'Metabuli', 'Kraken_0.1','MMSeqs2', 'Qiime2', 'TNT')) |>
left_join(groups_df, by = c('Database', 'Type')) |>
ggplot(aes(x = Type, y = `Alpha diversity index`, fill=Type, group = Type )) +
geom_boxplot(outlier.shape=NA) +
geom_point(aes(color=Type),
position = position_jitterdodge(jitter.width = 0.2),
alpha=0.5,
show.legend = FALSE) +
facet_wrap(~Database) +
geom_text(aes(x = Type, y = max(`Alpha diversity index`) + 2, label = groups),
#col = 'black',
size = 4) +
#coord_flip() +
theme_minimal() +
theme(axis.text.x = element_text( angle = 90, hjust = 1)) +
guides(fill="none") +
facet_wrap(~gene+Database, nrow = 3)
both_fig
my_save_plot(both_fig, 'exclusions_fig')
Let’s make the tables too:
tw <- twelve_exclusions_split_averaged
tw$Gene <- '12S'
six <- sixteen_exclusions_split_averaged
six$Gene <- '16S'
co <- co1_exclusions_split_averaged
co$Gene <- 'CO1'
both <- rbind(tw, six, co) |> relocate(Gene)
both |> my_save_table('all_exclusion_databases_scores')
both |> group_by(Gene, Database) |> arrange(desc(f1)) |> slice(1) |> ungroup() |>
select(Gene, Type, Database, accuracy, Precision, Recall, f1, f0.5, accuracy) |>
my_save_table('averaged_top1_exclusion_database_scores')